Centralized Dynamic Point Blanking in

LTE-Advanced Network for Inter-Cell Interference

Mitigation

Zhilan Xiong∗ , Min Zhang+ , Hakon Helmers , Matthew Baker+ , Philippe Godin and Dong Li∗
∗
Bell Labs China, Alcatel-Lucent Shanghai Bell, Shanghai, P. R. China, 201206
+
Alcatel-Lucent, Swindon, UK, SN5 7DJ

Alcatel-Lucent, Villarceaux, France, 91620
Email: [email protected], {min.hs.zhang, hakon.helmers,
matthew.baker, philippe.godin}@alcatel-lucent.com, [email protected]

Abstract—Dynamic Point Blanking (DPB) is a Coordinated also exploits the spatial domain for inter-cell interference mit-
MultiPoint (CoMP) technology that has attracted significant igation. However, DPB may provide higher gain than CS/CB
interest for inter-cell interference mitigation in wireless commu- in some scenarios, for example in a network with non-ideal
nication systems due to its potential performance gain, robustness backhaul (i.e. interconnection between transmission points),
and relatively low complexity. This paper describes the principle where the benefit of spatial domain coordination vanishes
of CoMP DPB as well as the standardized support that exists
quickly with backhaul delay, or with uneven traffic where some
for it in Long-Term Evolution Advanced (LTE-A) Releases 11
and 12, especially for centralized CoMP DPB; unified formulas lightly loaded cells can sacrifice some resources for the benefit
and solutions are provided for the optimization problems in of the network as a whole.
centralized CoMP DPB, together with corresponding evaluation CoMP DPS dynamically switches data transmissions to
results. According to these system-level evaluation results, it can
one UE between multiple points. DPS is usually used in
be concluded that CoMP DPB can provide a clear performance
gain in heterogeneous networks, is not sensitive to backhaul delay conjunction with DPB, where one cell-edge UE in a region
and has the potential to be used for both single-vendor and multi- of overlapping coverage from different transmission points
vendor CoMP deployments. can receive its data from the point with highest instantaneous
received power and muting is applied to the point(s) with
second and/or third highest instantaneous received powers in
I. I NTRODUCTION order to mitigate the interference to this cell-edge UE [9][10].

In cellular orthogonal frequency division multiple access The signaling enhancements in LTE-A Rel-11 for CoMP
(OFDMA) systems with frequency reuse of one, the throughput focus on the air interface between the base station (eNB) and
of the cell-edge user equipment (UE) is sensitive to the inter- UE and assume intra-eNB operation or a proprietary interface
cell interference from adjacent transmission points in the same between coordinated eNBs. In order to support multi-vendor
frequency channel. In order to mitigate this problem, coor- CoMP with a standardized interface between coordinated
dinated multipoint (CoMP) transmission/reception has been eNBs, LTE-A Rel-12 enhances the standardized X2 interface
considered in the 3rd Generation Partnership Project (3GPP) between eNBs, considering especially the case of non-ideal
Long-Term Evolution Advanced (LTE-A) standard [1][2]. In backhaul with non-zero latency.
a CoMP network, multiple transmission points coordinate This paper focuses on centralized CoMP DPB to study its
their data transmissions with the aim of improving cell-edge potential benefit in certain scenarios, proposes unified formulas
throughput and spectrum efficiency. for the optimization problems of CoMP DPB in LTE-A Rel-
11/Rel-12, and provides corresponding evaluation results.
The CoMP solutions supported in LTE-A Rel-11 for
downlink transmission include dynamic point blanking (DPB), The paper is organized as follows. The principle of CoMP
dynamic point selection (DPS) and coordinated scheduling and DPB is introduced in Section II. The optimization problems
coordinated beamforming (CS/CB) [1][3][4]. With the purpose and corresponding solutions for CoMP DPB in LTE-A Rel-
of inter-cell interference mitigation, CoMP DPB allows dy- 11/Rel-12 are presented in Section III and IV, respectively.
namic point muting in the time and frequency domains [5], Section V provides and analyzes corresponding system-level
while CS/CB coordinates transmission power and/or beam- evaluation results. The conclusions are drawn in Section VI.
forming to reduce the inter-cell interference [6-8]. Both DPB
and CS/CB need channel state information (CSI) of multiple II. P RINCIPLE OF C O MP DPB
transmission points to be available at each transmission point,
but do not require data sharing between coordinated points. As one of the more relevant inter-cell interference mitiga-
tion solutions in 3GPP LTE-A, CoMP DPB allows dynamic
CS/CB is expected to give greater benefit than CoMP DPB muting of transmission points in the time and frequency do-
in scenarios with higher traffic load, because of the fact that CB mains based on short-term CSI feedback from the UEs. In the

978-1-4799-8088-8/15/$31.00 ©2015 IEEE

 for each CSI process, the UE can be configured to measure
the signal power from multiple transmission points. By the
combination of muting and data transmissions from different
points, the network can configure a specific CSI-IM with
a suitable interference hypothesis for each UE to measure.
A Rel-11 UE can be configured with up to 3 CSI-IMs for
interference measurement [12], each of which can be measured
in up to two configurable subframe subsets, thus allowing one
Rel-11 UE to support up to 6 interference hypotheses for CSI
measurement/reporting. Therefore, for a victim UE needing
inter-cell-interference protection from the two strongest in-
terfering cells (for example point#1 and point#2), the eNB
Fig. 1. Principle of CoMP DPB. could configure the following four interference measurement
hypotheses for a given UE:
scenario shown in Fig. 1, UE#1 is served by point#1 but might • Hypothesis#1: The inter-cell interference with both
suffer strong interference from point#2; in order to protect point#1 and point#2 muting
the transmission quality of victim UE#1 using CoMP DPB,
point#1 and point#2 need to coordinate the data transmissions • Hypothesis#2: The inter-cell interference with point#1
to UE#1 by dynamically muting point#2 in the time-frequency muting and point#2 transmitting data to other UEs
transmission resources used by point#1. However, even when • Hypothesis#3: The inter-cell interference with point#1
UE#1 is scheduled, it may nonetheless be possible for point#2 transmitting data to other UEs and point#2 muting
to transmit data to its own UEs in some or all of the resources
used by point#1 for transmission to UE#1, depending on • Hypothesis#4: The inter-cell interference with both
factors such as traffic load, proportional fairness and overall point#1 and point#2 transmitting data to other UEs.
throughput impact. In order to provide sufficient information
to make the muting decisions, it is therefore helpful if each B. Muting Pattern Selection and UE Scheduling
potential victim UE feeds back multiple CSI measurements
representing different inter-cell interference hypotheses. According to received CSI measurements corresponding to
different interference hypotheses (i.e. muting patterns) as well
CoMP DPB can be classified into centralized DPB and as a predefined utility function, a DPB central coordinator at
distributed DPB, according to the entity making the decisions the network side can select the muting pattern that has the
on the muting pattern. In centralized DPB, the coordinated highest utility value for the transmission points that it controls.
points exchange information with their central coordinator. The In real implementations, the utility function may take into
central coordinator is responsible for the decisions of coordi- account quality of service (QoS) parameters like guaranteed
nated muting and/or coordinated scheduling in the CoMP clus- bit-rate and latency. However, for illustration purposes we
ter controlled by it. The individual transmission points collect consider here the maximal weighted global throughput (within
information and forward it to the coordinator, and implement a given cluster of coordinated transmission points) as the
the decisions made by the coordinator. In distributed DPB, objective function for selecting the muting pattern.
information is exchanged directly between transmission points.
Each transmission point makes and implements its own muting With the assumption of using single-user multiple-input
and scheduling decisions. multiple-output (SU-MIMO) as the scheduling scheme for
each coordinated point, the optimization problem for muting
In general, the optimal muting decisions for one trans- pattern selection over specific time-frequency resources could
mission point will depend on the muting decisions of its be formulated as:
neighbors. Therefore, centralized DPB has a higher decision-
making efficiency and lower implementation complexity than M

the distributed structure. max Rcell(m) |f (1)
f ∈F
m=1
III. C O MP DPB IN LTE-A R EL -11
where
A. CSI Measurement 
maxi∈Km wi RUE(i) |f , if cell m un-muting
Rcell(m) |f =
LTE-A Rel-11 introduced the concept of CSI processes to 0, otherwise
support feedback of CSI of multiple transmission points from
one UE [11]. One CSI process provides CSI measurement F denotes the set of available muting patterns for selection,
reports relating to one transmission point, where the channel M denotes the total number of coordinated cells in the CoMP
part of the CSI is measured on a non-zero power channel state cluster, Rcell(m) |f denotes the maximal weighted throughput
M
information reference signal (NZP CSI-RS) and the interfer- of cell m with muting pattern f , maxf ∈F m=1 Rcell(m) |f
ence is measured on a channel state information interference denotes the maximal weighted global throughput of the con-
measurement (CSI-IM) resource. The NZP CSI-RS comprises sidered CoMP cluster, Km denotes the set of active UEs served
coded reference signals from the transmission point, distributed by cell m, wi denotes a weighted coefficient of UE i (for
sparsely over the time-frequency resources in order to keep the example, related to the proportional fairness metric for the
overhead low. Using independent NZP CSI-RS configurations UE), and RUE(i) |f denotes the expected throughput of UE i
corresponding to muting pattern f (which could be derived
from the corresponding CSI reported by UE i).
If multi-user multiple-input multiple-output (MU-MIMO)
is used for UE scheduling at each transmission point, then the
optimization problem becomes:
M

max Rcell(m) |f (2)
f ∈F
m=1

where
 
max i∈Km wi RUE(i) |f , if cell m un-muting
Rcell(m) |f =
0, otherwise
Fig. 2. Inter-eNB CoMP DPB in LTE-A Rel-12.
The simplified optimization problems in (1) and (2) could
be solved by greedy search where the UEs are selected one by
one and only UEs that contribute most to improving the CoMP
assumed that the operation and maintenance (OAM) system
clusters weighted throughput are selected for scheduling.
is responsible for these configurations as shown in Fig. 3.

IV. C O MP DPB IN LTE-A R EL -12 According to the pre-configured NZP CSI-RS and CSI-IM
information from OAM, each local eNB can configure suitable
In order to support multi-vendor CoMP, LTE-A Rel-12 in- CSI processes for the CSI measurements at the UE side. For
cludes X2 interface enhancements for inter-eNB CoMP taking example, if the central coordinator requests CoMP Information
non-ideal backhaul into account. The enhanced X2 procedures from a local eNB, then the local eNB can configure corre-
include a Load Indication procedure for CoMP Information sponding CSI measurement/reporting from the UEs according
trigger and exchange, a Resource Status Reporting Initiation to pre-configured NZP CSI-RS and CSI-IM information; if
procedure, and a Resource Status Reporting procedure for the central coordinator stops the request of CoMP Information
periodic RSRP reporting [13]. The CoMP Information includes from a local eNB, then the local eNB can reconfigure the
a list of CoMP hypothesis sets, where each CoMP hypothesis CSI measurement/reporting in a way that is more suitable for
set is a collection of CoMP hypothesis(es) of one or more cells uncoordinated transmissions from a single transmission point.
and each CoMP hypothesis set is associated with a benefit
metric which indicates the utility of the hypothesis set. A
typical example of inter-eNB CoMP information exchange in B. Benefit Calculation
LTE-A Rel-12 is provided in Fig. 2, where: The benefit metric in LTE-A Rel-12 is defined as the benefit
expected to be obtained by one cell due to its associated CoMP
• Step 1: The UE reports its CSI measurements to its hypothesis set over some associated time-frequency resources
serving eNB; [14]. In inter-eNB DPB, one cell needs to report one or more
• Step 2: According to received CSIs from the UEs, the benefit metrics, and one benefit metric corresponds to one
eNB calculates the cell-specific benefit for each CoMP time-frequency resource indication and one muting pattern
hypothesis set and then sends information relating to including a set of muting information for transmission points
all cell-specific benefits and associated CoMP hypoth- involved in the CoMP cluster. We provide in this section two
esis sets to its central coordinator; examples of definitions for a benefit metric which could be
derived from CSI feedback from the UEs:
• Step 3: According to the received benefit information,
the central coordinator decides the muting pattern for Definition 1: This definition is based on normalized cell
each controlled eNB and informs the respective eNBs; throughput with and without inter-cell interference protection
from other cells:
• Step 4: According to the received muting pattern infor-
mation as well as the most recent CSI reports from the
UEs, each local eNB schedules data transmissions to
the UEs, including UE selection, precoding and MCS
selection.

Detailed solutions for CSI measurement, benefit calcula-

tion, muting pattern selection and UE scheduling are provided
in the following parts of this section.

A. CSI Measurement
In LTE-A Rel-12, the exchange of NZP CSI-RS and CSI-
IM configuration information between eNBs for CSI reporting
in CoMP DPB has not been standardized, and it may be Fig. 3. NZP CSI-RS and CSI-IM configuration information.
  100 f over the considered time-frequency resources. Similar to
Rcell(m) |f problem (1) and (2), the optimization problem in (7) could also
Bcell(m) |f = (3)
Rnorm 0
be solved by greedy search, where cells that are not muted
are selected one by one; cells which result in a throughput
where Bcell(m) |f denotes the normalized benefit expected to improvement for the CoMP cluster as a whole are allowed
be obtained by cell m corresponding to muting pattern f to transmit, and all others are muted. The frequency-domain
for certain time-frequency resources, Rcell(m) |f denotes the granularity for muting pattern selection at the central coordi-
maximal weighted throughput of cell m within the muting nator depends on the granularity of received benefit values,
pattern f over the time-frequency resources, Rnorm is a and could be subband-level or wideband-level.
parameter for benefit metric normalization and need to be pre-
configured at the network side, and
⎧ D. UE Scheduling
⎨a, x > a The central coordinator decides the muting patterns for
a
[x]b = x, b ≤ x ≤ a . (4) its controlled eNBs using the collected benefit information
⎩
b, x < b from these eNBs. The benefit information is cell-specific and
derived by each local eNB according to their own scheduling
The definition of Rcell(m) |f could refer to (1) for SU-MIMO
schemes. This mechanism, which is supported by LTE-A Rel-
and (2) for MU-MIMO.
12, maximizes the authority of local eNBs for their own UE
Definition 2: If considering the normalized throughput scheduling. In other words, the local eNBs in Rel-12 can select
gain obtained by one cell due to inter-cell interference pro- their preferred UE scheduling schemes, including whether SU-
tection from other cells as positive benefit, and the normalized MIMO and/or MU-MIMO are used, by themselves without
throughput loss obtained by one cell applying a muting pattern limitation from the central coordinator. Thus each eNB can
on its own resources in the case without inter-cell interference perform its own UE scheduling within the muting constraints
protection from other cells as negative benefit, then the benefit provided by the central coordinator, without considering the
metric could be defined as UE scheduling of its neighboring eNBs.

∗
100 V. S IMULATION AND E VALUATION
Rcell(m) |f − Rcell(m)
Bcell(m) |f = (5)
Rnorm In order to evaluate the performance of the schemes
−100 described above, this section provides numerical system-level
∗ ∗
where Rcell(m) = maxi∈Km wi RUE(i)  for SU-MIMO schedul- simulation results. The simulation platform is a 3GPP LTE-
∗ ∗
ing at cell m and Rcell(m) = max i∈Km wi RUE(i) for MU- Advanced system-level simulator, with detailed assumptions
∗ given in Table I.
MIMO scheduling at cell m, RUE(i) denotes the throughput
of UE i without inter-cell interference protection and can be The evaluation results with ideal backhaul are provided
derived by corresponding CSI feedback from UE i. in Table II, where the performance of SU-MIMO without
If one cell is muted over specific time-frequency resources, CoMP is used as the baseline for comparison. The evaluated
then there will be no throughput obtained by this cell over CoMP DPB scheme has coordination within one macro cell
those resources. Definition 2 in (5) could therefore be extended and the small cells within its geographical area. SU-MIMO
as follows: is supported for UE scheduling at each coordinated cell. The
⎧ 100 maximum number of SU-MIMO data streams per UE is 2, and
⎪ ∗
⎨ Rcell(m) |f −Rcell(m)
Rnorm , if cell m un-muting dynamic rank adaptation is supported. The weight for each
Bcell(m) |f = R∗ 100 0 UE in (1) is derived from the proportional fairness metric.
⎪
⎩− cell(m) , otherwise The cell-edge throughput of the UEs is defined as the 5th
Rnorm
0 percentile of the cumulative distribution function (CDF) of
(6)
where if cell m is not muted, then the network is ex- the user throughput. As shown in Table II, the performance of
pected to obtain additional (weighted) throughput gain of CoMP DPB with ideal backhaul and coordination within one
∗
Rcell(m) |f −Rcell(m) 100
macro cell and its covered small cells can give 48% cell-edge
[ Rnorm ]0 by CoMP DPB; if cell m is muted, then throughput gain with the cost of 1% cell-average throughput
the∗ network might lose the (weighted) throughput of at least loss compared to SU-MIMO. For the macro-cell UEs, there are
R
[ Rcell(m)
norm 0
]100 . 1% cell-average throughput gain and 20% cell-edge throughput
gain. For the small-cell UEs, there is 46% cell-edge throughput
C. Muting Pattern Selection gain, at the cost of 1% cell-average throughput loss.

The selection of the muting pattern with the criterion of Table III provides evaluation results for inter-eNB CoMP
global (CoMP cluster) benefit maximization in the central with non-ideal backhaul. The evaluated inter-eNB CoMP DPB
coordinator could be described as is between one macro cell and the small cells within its
M
geographical area. Each coordinated cell uses SU-MIMO for
 UE scheduling. The backhaul between eNBs is non-ideal with
max Bcell(m) |f (7) configurable transmission delay. CoMP DPB with 4ms two-
f ∈F
m=1
way backhaul delay is used as the baseline for performance
where Bcell(m) |f denotes the received benefit which is expected analysis. As shown in Table III, there is almost no perfor-
to be obtained by cell m with the assumption of muting pattern mance loss with the backhaul delay from 4ms to 16ms and
 TABLE I. S IMULATION PARAMETERS TABLE III. P ERFORMANCE OF DPB WITH NON - IDEAL BACKHAUL

Parameters Values used for evaluation Gain over baseline

Two-way backhaul delay
Cell average SE Cell-edge SE
Deployment scenario Heterogeneous network of macro and
small cells 4ms (baseline) 0% 0%
Cell structure Macro cell: hexagonal grid (wrap 8ms 1% -1%
around), 7 macro sites and 3 sectors 12ms 0% 3%
per site. Small cell: clusters randomly 16ms 0% 1%
dropped with uniform distribution with-
in macro geographical area; small cells 20ms -1% -4%
randomly dropped with uniform distri- 40ms -3% -4%
bution within cluster area; 10 small cells 60ms -3% -5%
per macro cell geographical area.
Number of UEs 630 UEs
Duplex mode FDD
System bandwidth 10 MHz shown clear performance gains for cell-edge throughput in a
Network synchronization Synchronized heterogeneous network. These gains are shown to be robust
Antenna configuration 2 Tx, 2 Rx, cross-polarized antennas
against two-way backhaul delays of up to around 16ms.
Antenna pattern 3D for macro cell and 2D omni-
directional for small cell R EFERENCES
Antenna tilt 15deg [1] 3GPP TR 36.819, “3rd generation partnership project; Technical specifi-
Feedback granularity in 5 PRBs (900kHz) cation group radio access network; Coordinated multi-point operation for
frequency domain LTE physical layer aspects (Release 11),” vol.11.2.0, September 2013.
Channel estimation Non-ideal [2] 3GPP TR 36.874, “3rd generation partnership project; Technical specifi-
Receiver type MMSE receiver cation group radio access network; Coordinated multi-point operation for
LTE with non-ideal backhaul (Release 12),” vol.12.0.0, December 2013.
DL overhead assumption Fixed, 0.3063
[3] S. Sun, Q. Gao, Y. Peng and Y. Wang, “Interference management through
Traffic model Full buffer CoMP in 3GPP LTE-advanced networks,” IEEE Wireless Communica-
Backhaul assumptions Infinite capacity; Zero latency for ideal tion, vol. 20, no. 1, pp. 59-66, February 2013.
backhaul and {4, 8, 12, 16, 20, 40, [4] D. Lee, H. Seo, B. Clerckx, E. Hardouin, D. Mazzarese, S. Nagata
60}ms latency for non-ideal backhaul. and K. Sayana, “Coordinated multipoint transmission and reception in
Link adaptation Non-ideal LTE-advanced: deployment scenarios and operational challenges,” IEEE
Communication Magazine, vol. 50, no. 2, pp. 148-155, February 2012.
[5] M. Li, X. Yun, S. Nagata, L. Chen, “Power allocation of dynamic
TABLE II. P ERFORMANCE OF DPB WITH IDEAL BACKHAUL point blanking for downlink CoMP in LTE-advanced,” International
Conference on Wireless Communications & Signal Processing, 24-26
Cell average SE Cell-edge SE Oct. 2013, pp. 1-5.
UE type Schemes [6] Z. Xiong, H. Yang, M. Zhang, Y. Meng and Y. Pan, “Feedback and
(bps/Hz/cell) (bps/Hz/UE)
scheduling for coordinated beamforming of CoMP in LTE-advanced
SU-MIMO 17.03 0.082 system,” IEEE 78th Vehicular Technology Conference (VTC Fall), 2013,
All UEs CoMP DPB 16.84 0.121 pp. 1-5
Gain -1% 48% [7] U. Jang, H. Son, J. Park and S. Lee, “CoMP-CSB for ICI nulling with
SU-MIMO 2.15 0.061 user selection, IEEE Transactions on Wireless Communications,” vol. 10,
no. 9, pp. 2982-2993, September 2011.
Macro UEs CoMP DPB 2.17 0.073
[8] Q. Li, Y. Yang, S. Fang and G. Wu, “Coordinated beamforming in
Gain 1% 20% downlink CoMP transmission system,” in Proc. International ICST
SU-MIMO 1.49 0.087 Conference on Communications and Networking in China, 25-27 August
Small-cell UEs CoMP DPB 1.47 0.127 2010, pp. 1-5.
Gain -1% 46% [9] M. Feng, X. She, L. Chen and Y. Kishiyama, “Enhanced dynamic cell
selection with muting scheme for DL CoMP in LTE-A,” IEEE Vehicular
Technology Conference, 16-19 May 2010, pp. 1-5
[10] G. Morozov, A. Davydov, and I. Bolotin, “Performance evaluation of
dynamic point selection CoMP scheme in heterogeneous networks with
only marginal performance degradation for the backhaul delay FTP traffic model,” The 4th International Congress on Ultra Modern
from 20ms to 60ms. According to these results, it could be Telecommunications and Control Systems and Workshops (ICUMT),
concluded that the solutions in section IV are robust against 2012, pp. 922-926
two-way backhaul delays up to around 16ms and are suitable [11] 3GPP TR 36.213, “3rd generation partnership project; Technical spec-
for inter-eNB CoMP with a common X2 interface in LTE-A ification group radio access network; Evolved universal terrestrial radio
Rel-12. access (E-UTRA); Physical layer procedures (Release 11),” vol.11.7.0,
June 2014, pp. 55-88
[12] 3GPP TR 36.331, “3rd generation partnership project; Technical spec-
VI. C ONCLUSION ification group radio access network; Evolved universal terrestrial radio
access (E-UTRA); Radio resource control (RRC); Protocol specification
We have described the principle of CoMP DPB and stan- (Release 11),” vol.11.8.0, June 2014, pp. 191-193
dardized support for it in LTE-A Rel-11 and Rel-12. We have [13] Samsung, “Way forward on WI: inter-eNB CoMP for LTE,” R3-141488,
proposed unified formulas and corresponding solutions for the 3GPP TSG-RAN WG3 #84, Seoul, Korea, 19-23 May 2014.
optimization of transmission point muting patterns with both [14] Samsung, “Introduction of inter-eNB CoMP,” R3-141487, 3GPP TSG-
ideal and non-ideal backhaul. System-level simulations have RAN WG3 #84, Seoul, Korea, 19-23 May 2014.